Split getTestRecordId only on the first occurence of '#'

When the first time running the CtsLibcoreTestCases, the tests' names
in the test_result.xml are right while the tests' names stored in
test-record.pd are missing the suffixes "#Cyrl" and "#Latn". For
example, the full test name is "libcore.highmemorytest.java.text.
DateFormatTest#test_toPattern_allLocales[az__#Cyrl]", and the
mutilated name is "libcore.highmemorytest.java.text.DateFormatTest#
test_toPattern_allLocales[az__ ". Therefore, when retrying this module,
the cts load from the wrong test-record.pd, miss 819 tests and add 486
mutilated tests. The delta between running and collecting is 333.

When generating pd file, the tradefed will first generate 3 tmp
files: test-record.pd0, test-record.pd1 and test-record.pd2. Among
them, test-record.pd1 contains the test names. When finishing tests,
tradefed will read tmp files and process them to generate the final
pd file. When tradefed processing record, it will split the
childProto's TestRecordId with "#" and only get info[0] and info[1],
which results in the loss of the suffix string after the second “#”.
To fix this bug, tradefed should split getTestRecordId only on the
first occurrence of '#'.

Test: run retry. All tests in CtsLibcoreTestCases should be run.
Bug: 344688732
Change-Id: I0b53fb401da1dce73f0ce4540f5e8312646f9748
Signed-off-by: Cui, Yuxin <yuxin.cui@intel.com>
Deng, Bing <bing.deng@intel.com>

• src/com/android/tradefed/result/proto/ProtoResultParser.java[diff]

1. Create your change in Gerrit
2. Add the reviewer named "Tradefed Codereview" (email: tradefed-codereview@tradefederation.google.com.iam.gserviceaccount.com)
3. Review the code review guidance at go/tf-guidelines and go/tradefed-code-reviews
4. GWSQ should add a couple of people from the team to review your code and give feedback.